DEV Community

Zane Milakovic
Zane Milakovic

Posted on

How we built cross-region uptime verification (and why single-location monitoring is broken)

If you've ever been woken up at 3am by a monitoring alert that turned out to be nothing, you already understand the problem.

Most uptime monitoring works like this: a server in Virginia pings your site every minute. If it gets a bad response, it sends you an alert. Simple, effective, and wrong about 20% of the time.

That number isn't made up — it's roughly what I saw across my own client sites over two years of using various monitoring tools. About one in five alerts was a false positive caused by network routing issues, transient DNS problems, or a brief hiccup at the monitoring provider's own data center.

The fix is obvious (in hindsight)

When a check fails, don't immediately alert. Instead, trigger verification checks from other regions. If Chicago says your site is down but Amsterdam, Virginia, and Singapore all say it's fine — that's not an outage. That's a network blip.

This is what I built into FlareWarden. Here's roughly how it works:

Step 1: Initial check fails. One of our 18 monitoring regions reports a failure. Timer starts.

Step 2: Cross-region verification. We immediately fire checks from multiple other regions. The number of confirming regions required is configurable — you might want 2 out of 3 for a personal project, or 4 out of 5 for production infrastructure.

Step 3: Consensus determines outcome. If verification checks confirm the outage, alert fires. If they don't, we log it as a regional issue and move on. You sleep through the night.

The whole verification loop typically completes in 30-60 seconds. Fast enough to catch real outages quickly, slow enough to filter out noise.

The parent/child thing

The other architectural decision I'm pretty happy with is the monitor hierarchy.

Traditional monitoring gives you a flat list: site A is up, site B is down. But that's not how web apps actually work. Your e-commerce site depends on Stripe for payments, Cloudflare for CDN, maybe Shopify for inventory. When Stripe goes down, your site isn't down — checkout is broken but the rest works fine.

FlareWarden uses a parent/child model. Your main site is the parent monitor. Dependencies like Stripe or your CDN are child monitors of type "dependency." When a dependency fails, the parent status changes to "degraded" rather than "down." Content monitors (checking that specific text exists on a page) are independent children — they alert you directly without affecting the parent status.

This means your status page automatically reflects what's actually happening: "Website operational, payment processing degraded" is way more useful than "Website down."

Auto-discovery

The last piece I wanted to get right was setup friction. Configuring monitors manually for every domain, every SSL cert, every third-party integration — it's tedious and you always miss something.

FlareWarden's Smart Setup scans your URL and automatically identifies: third-party service dependencies (we recognize 700+ services), SSL certificate details and expiry dates, critical page content that should always be present, and your overall tech stack.

You paste a URL, review what it found, and click confirm. Full monitoring in about two minutes.

If you want to try it

Free tier is 15 monitors with 5-minute checks, no credit card, no expiry. I'm running founding member pricing (40% off forever) on paid plans through June if you want faster checks or more monitors.

https://flarewarden.com

Would genuinely love technical feedback from this community. The verification logic, the monitor hierarchy, the auto-discovery — poke holes in any of it. That's how it gets better.

Top comments (0)